Prediction of students´ academic performance: Adapt a methodology of predictive modeling for a small sample size

The present study adapted the methodology illustrated in a published study for a small sample size [1]. A total of 48 students who were enrolled in an engineering dynamics course were involved in the present study. Multiple linear regression (MLR) and support vector machine (SVM) models were developed and tested. Three criteria were defined to assess the accuracy of the predictive models. The criteria were the number of accurate predicted grade ranges, missing alarm, and maialami. Results indicated that 1) the methodology for developing the predictive models could be employed to different institutions of higher education; and 2) the SVM model may have higher prediction accuracy than the MLR model when the sample size is small.